vg 1.50.1 - Monopoli

Don't forget to mark the static binary executable:

chmod +x vg

Docker Image: quay.io/vgteam/vg:v1.50.1

Buildable Source Tarball: vg-v1.50.1.tar.gz

Includes source for vg and all submodules. Use this instead of Github's "Source Code" downloads; those will not build as they do not include code for bundled dependencies that the vg build process needs.

This release includes:

• vg autoindex --workflow map can index GFAs with many W lines
• vg autoindex -w map and -w mpmap won't enter an infinite loop when they can't write to disk
• GRCh38#chr1 style path names in GFA P lines should now be parseable again

Updated Submodules

• gcsa2
• libhandlegraph

vg 1.50.0 - Monopoli

Don't forget to mark the static binary executable:

chmod +x vg

Docker Image: quay.io/vgteam/vg:v1.50.0

Buildable Source Tarball: vg-v1.50.0.tar.gz

Includes source for vg and all submodules. Use this instead of Github's "Source Code" downloads; those will not build as they do not include code for bundled dependencies that the vg build process needs.

This release includes:

• CI test jobs now cache pulled Docker images
• GAF output should now have more correct path end positions and block lengths
• Paths that look like PanSN but aren't, due to having a non-numeric haplotype number field, will no longer be parsed, and should thus no longer produce crashes due to parsing failures.
• Haplotype sampling now copies the vg node to GFA segment translation correctly from the original graph.
• vg minimizer requires a distance index for building a minimizer index.
• -S option added to vg call to select reference paths by sample name. This is more convenient as it allows, ex -S GRCh38 to be used in place of -p GRCh38#0#chr1 -p GRCh38#0#chr2 ... Such selection is necessary when the graph has more than one reference sample and vg call will now refuse to handle graphs with multiple reference samples unless paths are selected with -S or -p.
• vg filter can filter to only mapped or only unmapped reads
• vg deconstruct changed back to writing the full sample / hap/ contig name in VCF contig field. In order to just write the contig name (like in the past few versions of vg), use the new -C option.

Updated Submodules

• gbwtgraph
• gcsa2
• libhandlegraph
• libvgio

vg 1.49.0 - Peschici

Don't forget to mark the static binary executable:

chmod +x vg

Docker Image: quay.io/vgteam/vg:v1.49.0

Buildable Source Tarball: vg-v1.49.0.tar.gz

Includes source for vg and all submodules. Use this instead of Github's "Source Code" downloads; those will not build as they do not include code for bundled dependencies that the vg build process needs.

This release includes:

• Giraffe can now use weighted minimizers, which try to avoid selecting frequent kmers as minimizes.
• vg inject can read BAM files with unmapped reads
• vg giraffe now has --match, --mismatch, --gap-open, --gap-extend, and --full-l-bonus options to control alignment scoring.
• Fix crash during assertion in vg deconstruct on PGGB graph that was introduced in v1.48.0
• : character now allowed in path name during contig:start-end range extraction from command line options (ie in vg chunk).
• vg now builds with C++17 on Mac, as required by the version of Protobuf packaged in Homebrew
• vg now deduplicates arguments from pkg-config, to limit command line length with Protobuf's 30-odd Abseil dependencies.
• Better default parameters for haplotype sampling.
• vg clip crash on PackedGraphs fixed.
• Mac CI now collects Homebrew debug info
• vg's CI can now run on local Gitlab runners
• CI no longer does extra Docker builds without proper caching
• vg giraffe -b fast preset now works again and is under test
• Serialized mutable graphs keep proper track of the number of edges they contain

Updated Submodules

• gbwtgraph
• libbdsg
• libvgio

vg 1.48.0 - Gallipoli

Don't forget to mark the static binary executable:

chmod +x vg

Docker Image: quay.io/vgteam/vg:v1.48.0

Buildable Source Tarball: vg-v1.48.0.tar.gz

Includes source for vg and all submodules. Use this instead of Github's "Source Code" downloads; those will not build as they do not include code for bundled dependencies that the vg build process needs.

This release includes:

• vg chunk will now report an error if asked to chunk reads that do not go with the graph
• vg autoindex can construct linear reference indexes from a FASTA file
• vg gbwt will now refuse to add reference sample names with # in them, and will try and advise on what the tags are supposed to be like
• vg surject can project to paths that intersect themselves in the reverse orientation
• vg surject will now print warning messages when processing a read or pair takes a suspiciously long amount of time.
• vg giraffe should no longer try to put hypothetical sequencing errors in empty intervals, and should report errors in MAPQ cap computation in a more debuggable way.
• Crashes now include the stack trace by default; set VG_FULL_TRACEBACK=0 to suppress it to a file.
• vg surject and vg giraffe should now include relevant read name hints when crashing in many cases.
• Added crash_unless() as an alternative to assert() that reports these hints. We eventually want to use it everywhere.
• Crash reports now have cool hyperlinks.
• vg surject will limit itself to 200 anchors per target path segment by default; use the new -a/--max-anchors option to control this limit. Surjection against PGGB graphs may require --max-anchors 20 to complete.
• vg surject may be able to limit itself to considering only high-scoring surjections in some cases.
• vg construct now properly handles the case where it is looking for the end of an inversion from 1 base before it
• vg construct will no longer try and coalesce nodes at construction chunk boundaries when those nodes have alt paths that visit them or edges to their outside endpoints. This should fix some crashes and incorrect placement of structural variant breakpoints in the graph.
• Update vcflib to current version plus build and parser fixes
• vg construct should now be faster when variants are extremely long and overlap each other
• vg chunk now outputs PackedGraph instead of Protobuf by default (unless -T is used). Also, output files now get the .vg file extension for any non-GFA format (use vg stats -F to check the underlying format of any graph).
• Snarl clipping bug in vg clip fixed so that when there are multiple different reference traversals in a snarl (common in PGGB output), then none of them are chopped.
• Fixed build against Ubuntu 22.04's pybind11
• Docker containers now have /usr/bin/time for profiling

Updated Submodules

• htslib
• vcflib
• gbwtgraph
• libbdsg
• tabixpp

vg v1.47.0 - Ostuni

Don't forget to mark the static binary executable:

chmod +x vg

Docker Image: quay.io/vgteam/vg:v1.47.0

Buildable Source Tarball: vg-v1.47.0.tar.gz

Includes source for vg and all submodules. Use this instead of Github's "Source Code" downloads; those will not build as they do not include code for bundled dependencies that the vg build process needs.

This release includes:

• vg sim and vg stats -a sped up for GBZ input
• Giraffe now uses the watchdog to detect slow reads
• vg construct should no longer fail assertions and will instead report errors.
• vg construct now handles IUPAC codes in the reference as Ns even if they are covered by symbolic structural variants
• Faster haplotype sampling with vg haplotypes.
• vg stats -a also outputs statistics on alignment scores and mapping quality.
• vg giraffe should no longer crash if the distance index is read-only.
• vg rna now supports the GBZ format for the input graph and haplotypes (new option --gbz-format).
• vg convert now defaults to PackedGraph instead of HashGraph if no output format selected.
• New option vg clip -s to remove stubs (dangling nodes not on ref path)
• vg call and vg deconstruct now only apply node ID translation from GBZ inputs if new -O is used.
• vg surject will now enforce that the reads it is surjection actually were mapped against the graph you are surjecting against. Right now it checks node IDs and lengths. You can turn this off with -V/--no-validate.
• vg gbwt now accepts a -I/--gg-in option, which lets you load a .gg file and a .gbwt file and combine them into a .gbz graph.
• vg validate now accepts a -A/--gam-only option which will validate only the provided alignment's agreement with the graph, and not the graph itself.
• The vg surject/vg giraffe error: couldn't identify a path corresponding to surjected read error message has been improved to dump more information about the offending read and path.
• When selecting paths to surject to, a warning will now be printed if the user asks for a path with a []-enclosed subrange at the end. The base path name without the [] subrange coordinates should usually be used instead, because that is the space in which the SAM/BAM output will have its coordinates specified.
• The vg surject Graph does not have a path named error message should now no longer print pointer values, and is extended to explain a bit more about subpaths.

Updated Submodules

The kff-cpp-api and libbdsg submodules have been updated.

vg 1.46.0 - Altamura

Don't forget to mark the static binary executable:

chmod +x vg

Docker Image: quay.io/vgteam/vg:v1.46.0

Buildable Source Tarball: vg-v1.46.0.tar.gz

Includes source for vg and all submodules. Use this instead of Github's "Source Code" downloads; those will not build as they do not include code for bundled dependencies that the vg build process needs.

This release includes:

• Long read Giraffe codepath now falls back to non-GBWT alignment for very long tails, which is slow but at least tends to finish
• Long read Giraffe codepath refuses to use Dozeu for tails, because the tails are very long and Dozeu will clobber the stack when given a long alignment
• More knobs have been added to long read Giraffe to tweak what inter-pre-cluster connections are sent to reseeding, and what chains are actually made into alignments
• vg stats now reports on time-usage information in GAM files if available
• Wiki tutorial on programming with libbdsg and gbwtgraph is now under CI test.
• Rescue alignment in vg giraffe paired-end mode should no longer decide it rescued off of the wrong alignments
• GAMP files no longer lose the "secondary" annotation when converted to GAM
• New benchmarking and read-simulating scripts for testing long-read Giraffe
• Fixed a crash in Giraffe correctness tracking in the long-read codepath due to out-of-bounds accesses into previous stages in the funnel
• Reading SAM/BAM/CRAM files into a graph (i,e, vg inject) will now bail out and complain if they are against haplotypes and not reference or generic paths (because positional lookup is likely to be too slow to be practical)
• vg inject now defaults to the normal default number of threads
• vg gamcompare now has a -n/--rename options for comparing GAM files annotated with position on the same contigs but with different names.
• vg annotate now uses a ReferencePathOverlayHelper to make sure it has fast access to the positions of graph nodes along paths.
• vg CI now tests against sequenceTubeMap using its recommended Node version
• vg rna will no longer in certain cases skip the first line when the annotation input has a header
• vg rna no longer crashes when adding splice-junction from a BED file with introns
• scripts/make_pbsim_reads.sh now works with local graph files in addition to S3 URLs
• scripts/lr_benchmark.sh now downloads and uses CHM13 graphs
• Chaining lookback now stops at 15 total items max
• Tail alignment with GSSW now refuses to fill more than 16 mibi-cells
• Fix off-by-1 array size bug in vg clipedge clipping
• Faster vg chunk on GBZ input
• Highly experimental vg haplotypes subcommand for sampling haplotypes based on kmer counts.
• vg giraffe now preloads the distance index into memory before mapping any reads
• Makefile should deal better with protoc not being installed
• Surjecting now works for interleaved GAFs, even if one or both reads are unmapped.

Updated Submodules

The libbdsg and gbwtgraph submodules have been updated.

New Submodules

The kff-cpp-api submodule has been added.

vg 1.45.0 - Alpicella

Don't forget to mark the static binary executable:

chmod +x vg

Docker Image: quay.io/vgteam/vg:v1.45.0

Buildable Source Tarball: vg-v1.45.0.tar.gz

Includes source for vg and all submodules. Use this instead of Github's "Source Code" downloads; those will not build as they do not include code for bundled dependencies that the vg build process needs.

This release includes:

• vg clip -d now removes edges not meeting the depth threshold, in addition to nodes
• vg paths -C reports both undirected (visit any node twice) and directed (visit any node twice in same orientation) cyclicity
• Macports-based installation instructions have been simplified
• vg Mac build now works even if g++ --version doesn't actually have the version string on the first line
• vg no longer supports the "ODGI" format. The implementation that vg used was never compatible with the implementation in the actual odgi tool. If you have graphs in the old vg-odgi format, use vg 1.44.0 to convert them to another format. To exchange data with the odgi tool, use the GFA format.
• Fix stats -p option
• vg giraffe --align-from-chains invokes a separate (still experimental) long-read-optimized codepath.
• vg giraffe option parsing uses a new setup that makes it easier to add and report on options.
• Giraffe funnel explanations can track correctness and placed-ness along the length of a read.
• vg mpmap has increased sensitivity for detecting splice junctions
• vg autoindex can produce graphs (included spliced graphs) from FASTAs without requiring VCFs
• vg autoindex now splits the indexing for vg mpmap from the indexing for rpvg. The previous behavior can be recapitulated by indicating both workflows: --workflow mpmap --workflow rpvg
• vg chunk and vg trace can now get haplotypes from an input GBZ file
• vg chunk can use --no-embedded-haplotypes to ignore haplotypes from a GBZ.

Updated Submodules

The libbdsg and libhandlegraph submodules have been updated.

vg 1.44.0 - Solara

Don't forget to mark the static binary executable:

chmod +x vg

Docker Image: quay.io/vgteam/vg:v1.44.0

Buildable Source Tarball: vg-v1.44.0.tar.gz

Includes source for vg and all submodules. Use this instead of Github's "Source Code" downloads; those will not build as they do not include code for bundled dependencies that the vg build process needs.

This release includes:

• GAF output fixed to 1) no longer have consecutive deletions in the CIGAR and 2) be valid for split mappings within the same node (which can happen when mapping long reads with vg map)
• vg autoindex can auto-tune some key indexing parameters for increased robustness.
• vg autoindex no longer crashes on VCFs that contain no samples.
• Add the flag --num-bp-per-min to Giraffe to adjust the number of selected minimizers based on the read length.
• vg main GFA loader can now handle HPRC-style GFAs where the same path exists as rGFA tags and P lines.
• L-lines in GFA output changed back to have 0M cigars (reverting switch to * in v1.31.0)
• vg surject can annotate SAM/BAM records by all of the sequences it attempted to realign to
• vg surject has improved stability on cyclic graphs
• Handle GBWTGraphs and GBZ graphs that do not contain a translation correctly when a translation is needed.
• vg construct will no longer fail with an assertion error about last_edit_end != -1 and will instead report the variants that confused it.
• vg construct will skip over and warn about variants that do not actually change anything.
• vg should now only link one copy of Protobuf into the non-static build
• vg should now build against newer libomp from Homebrew which is keg-only
• vg rna now support reference transcript paths where exon boundaries are on opposite strands
• vg convert --drop-haplotypes will drop haplotype paths from the output graph.
• preliminary GBZ input support in vg deconstruct and vg call
• vg surject, vg giraffe, vg mpmap, vg filter, and vg deconstruct now accelerate paths-on-node queries with an overlay, so working from a GBZ will no longer be quite as slow relative to working from an XG
• libbdsg PackedReferencePathOverlay should no longer crash for short paths
• Update distance index to distance index 2 to be more efficient and make clustering faster
• Minimizers now have a payload with two ints
• DI2 files will need to be rebuilt as well as minimizers that use the distance index
• New vg chunk option -S can be used to extract every snarl that is fully contained in the the given path region (ie specified with -p). This can be used instead of context steps -c. The advantage is that -S will return everything inside the region and nothing outside the region (barring the start and end nodes), which helps with the problem of pulling out massive amounts of neighbouring regions when jacking up -c for complex subgraphs. The disadvantage is that if the region specified contains only parts of snarls, the results will be a misleadingly simple graph.
• vg viz no longer silently fails when asked to draw a PNG that is too big for Cairo

Updated Submodules

The gbwt, gbwtgraph, gcsa2, libbdsg, libhandlegraph, libvgio, and xg submodules have been updated.

vg 1.43.0 - Barisano

Don't forget to mark the static binary executable:

chmod +x vg

Docker Image: quay.io/vgteam/vg:v1.43.0

Buildable Source Tarball: vg-v1.43.0.tar.gz

Includes source for vg and all submodules. Use this instead of Github's "Source Code" downloads; those will not build as they do not include code for bundled dependencies that the vg build process needs.

This release includes:

• Mac CI now runs on MacOS 12
• uniform_int_distribution unit tests no longer hang on MacOS 12
• vg giraffe can now --align-from-chains to do long-read alignments
• Makefile now supports make bin/unittest/<test_file_name> to build a dynamically-linked binary for just one file of unit tests, for faster iteration.
• -c option added to vg depth path coverage in order to count each time a path steps on a position. By default vg depth continues to only count each path once

Updated Submodules

The vcflib submodule has been updated.

vg 1.42.0 - Obolo

Don't forget to mark the static binary executable:

chmod +x vg

Docker Image: quay.io/vgteam/vg:v1.42.0

Buildable Source Tarball: vg-v1.42.0.tar.gz

Includes source for vg and all submodules. Use this instead of Github's "Source Code" downloads; those will not build as they do not include code for bundled dependencies that the vg build process needs.

This release includes:

• vg releases should now have their Docker containers built by our CI.
• vg releases should now have ARM/amd64 multiarch Docker tags.
• README now documents using Giraffe.
• SAM/BAM alignment output and surjection now tries to use reference paths, and then generic paths, but won't use haplotype paths by default